This paper presents methodology to identify the surface roughness value in CNC machining process using a soft computing approach. The aim of this paper is to achieve a roughness accuracy value above 95% and reduce the error rate to below 5% by using an artificial neural network. An artificial neural network method was selected to improve the time of inspection. Fourier transformation method will be used to extract the turning workpiece image, which is the squared value of the major frequency and principal component magnitude. Primary machining parameters such as feed rate, depth of cut, speed, frequency range, gray scale value, and conventional measurement value feed are used as the training input in the artificial neural network. Based on the training sample, the artificial neural network generates the vision measurement value for the testing samples that is compared to the stylus probe measurement value to predict the error rate and accuracy. The novelty of this work is to create an effective methodology using artificial neural network techniques to detect surface roughness errors of materials used in manufacturing industries.

The objective of the present study was to predict the surface topological characteristics of Al–B₄C composite electrodes and oil hardened, non-shrinking (OHNS) die steel in the electrical discharge machining (EDM) process. The surface characteristics of the composite electrodes were evaluated using a scanning electron microscopy (SEM) and EDAX analysis. The surface roughness and hardness of the OHNS die steel were measured using a Stylus probe and Brinell hardness tester, respectively. The composite electrodes were prepared with Al 6063 and B₄C materials. In the stir casting process, the two materials were mixed at the molten state at different compositions. The chemical composition of the composite electrodes were analyzed by SEM and EDAX testing. The surface roughness of the OHNS steel was measured using a Brinell hardness tester. Based on SEM and EDAX results, 92% Al 6063, 8% B₄C produced the best surface roughness in OHNS die steel.

This paper proposes an efficient methodology for predicting surface roughness using different soft computing approaches. The soft computing approaches are artificial neural network, adaptive neuro-fuzzy inference system and genetic algorithm. The proposed surface roughness prediction procedure has the following stages as feature extraction from the materials, classifications using random forests, adaptive neuro-fuzzy inference system (ANFIS). In this paper, the statistical features are extracted from material images as skewness, kurtosis, mean, variance, contrast, and energy.The surface roughness accuracy value varied between ANFIS and random forest classification in every measurement sequence. This limitation can be overcome by the genetic algorithm to optimize the best results. The optimization technique can produce more accurate surface roughness results for more than 98% and reduce the error rate up to 0.5%.